On Thu, Mar 03, 2022 at 10:05:22AM -0700, Nathan Chancellor wrote:
Hi Naresh,
On Thu, Mar 03, 2022 at 09:09:56AM +0530, Naresh Kamboju wrote:
[Please ignore this email if it is already reported]
Linux next-20220302 mips builds failed [1] & [2].
Regressions found on mips:
- mips-clang-12-defconfig
- mips-clang-13-allnoconfig
- mips-clang-12-allnoconfig
- mips-clang-12-tinyconfig
- mips-clang-13-defconfig
- mips-clang-11-tinyconfig
- mips-clang-11-defconfig
- mips-clang-nightly-allnoconfig
- mips-clang-14-tinyconfig
- mips-clang-14-defconfig
- mips-clang-11-allnoconfig
- mips-clang-nightly-tinyconfig
- mips-clang-13-tinyconfig
- mips-clang-nightly-defconfig
- mips-clang-14-allnoconfig
make --silent --keep-going --jobs=8 O=/home/tuxbuild/.cache/tuxmake/builds/1/build LLVM=1 LLVM_IAS=1 ARCH=mips CROSS_COMPILE=mips-linux-gnu- 'HOSTCC=sccache clang' 'CC=sccache clang' defconfig
<snip>
fatal error: error in backend: Invalid register name global variable
Thanks a lot for the report! Our CI caught this as well, I have filed an issue on our tracker and the upstream LLVM one:
https://github.com/ClangBuiltLinux/linux/issues/1608 https://github.com/llvm/llvm-project/issues/54174
CC Kees, just as an FYI because of commit 200ed341b864 ("mips: Implement "current_stack_pointer"").
Oh! That's surprising. Thanks for digging out the details. The comment in the code is amusing:
Register MipsTargetLowering::getRegisterByName(const char *RegName, LLT VT, const MachineFunction &MF) const { // Named registers is expected to be fairly rare. For now, just support $28 // since the linux kernel uses it.
Anyway, I guess I'll need to send a fix to make this compiler-version dependent...